Dealing with percentages for variables

[Joe Tuckles]

In order to run a regression, do I need to first transford any variables that are percentages onto a logit scale?

[Rich Goldstein]

no

if they are really percentages, however, you probably want to divide by 100 so they become proportions and then use either -fracreg- or -betareg-; see

help fracreg
help betareg

[Joe Tuckles]

OK and what if some of the variables going into regression are not proportions and some are? Would I still use one of those regressions?
Thanks

[Rich Goldstein]

sorry, I assumed that the "percentage" variable was the outcome variable - the predictors/covariates can be anything you want; if your percentage variable is not the (an) outcome variable, just use whatever regression you would normally use in that situation and, yes, include your percentage variable as a covariate/predictor/independent variable

for further advice, you need to follow the FAQ and show example data, using -dataex-, and describe what you want from your regression

[Joe Tuckles]

Excellent thank you